Module 3 - Linear Models

Overview and Deliverables

This week we will cover linear regression, a topic you are already quite familiar with from previous classes. We will take a different perspective on this subject, contextualizing it within machine learning as the default, baseline model that should be used for regression and also as the most interpretable model, and also discussing it from the distinct goals of prediction and understanding. One conclusion will be that variable selection is very different depending on your goal.

You should begin working on Lab 2 and continue project brainstorming and group formation.

Learning Objectives

• Ordinary Least Squares for Regression
• Baseline Models
• Intrepreting Regression Coefficients, Confounders and Colliders
• Prediction versus Inference
• Weighted Least Squares
• Extending Regression using Feature Engineering and Nonlinear Transformations

Readings

• ISLP (Introduction to Statistical Learning): Chapter 3

Labs

• ISLP (Introduction to Statistical Learning): Chapter 3 Lab
• Ipython Notebook Version

Videos

• Section 3.1: Simple Linear Regression
• Section 3.2: Hypothesis Testing and Confidence Intervals
• Section 3.3: Multiple Regression
• Section 3.4: Some Important Questions
• Section 3.5: Extensions of the Linear Model

Coding Videos

• Statsmodels Linear Regression
• Statsmodels Multiple Regression
• Statsmodels Features